At the 2025 Australasian Fleet Education & Leadership Summit, Geotab’s Vice President of Data & Analytics, Mike Branch, delivered a powerful message for fleet professionals embracing AI: don’t overlook the basics. His keynote focused heavily on the importance of good quality, clean, and accurate data—without which the promises of artificial intelligence can quickly become a liability rather than a benefit.
Branch used humour and relatable analogies to make a serious point: “Garbage data, garbage insights,” he said. “It doesn’t matter what you do, and guess what? With agents, it’s getting worse. It only exemplifies that—if you have garbage data, you’re going to get terrible, terrible insights multiplied by a million.”
As fleets become more sophisticated, the use of machine learning, generative AI, and intelligent agents is increasing. From predictive maintenance and safety modelling to automated incident reporting and driver coaching, these systems rely on historical and real-time data to make informed decisions. But if the data fed into them is incomplete, outdated, or incorrect, the insights generated are not just misleading—they can actively cause harm.
Branch illustrated the risks of bad data with a fictional—but believable—example. “I might be driving actually quite well, and it’s telling me I’m slamming on the brakes, telling me I’m speeding. I’m going to turn that thing off in an instant,” he warned. In this scenario, a driver disengages from the system due to frustration. Worse still, poor data quality could lead an AI agent to reward the wrong behaviour or even alert the wrong emergency contact after a collision.
The consequences go beyond inconvenience—they erode trust in the technology. “The very important thing that we can do today is… make sure we have the tools to measure [our goals] appropriately and have quality data,” Branch explained. Even the most sophisticated AI platforms can’t compensate for flawed inputs. When the output is used to inform operational decisions, the cost of getting it wrong becomes tangible in downtime, safety incidents, or lost productivity.
So what can fleet managers do? First, ensure that telematics devices and sensors are properly calibrated and maintained. Next, build data hygiene into routine fleet processes—such as regular validation of vehicle configurations, driver assignments, and maintenance logs. Finally, make sure data definitions are clear and consistent across systems. As Branch asked, “What is downtime? What is utilisation? If you can’t measure every single piece, you’re going to have these agents go bananas.”
AI isn’t a magic wand—it’s a powerful tool that magnifies what it’s given. For fleet managers, the message is clear: if you want AI to work for your operation, you must start with clean, high-quality data.
“Don’t lose that focus,” Branch concluded. “Solve the data quality problem first.”
